Factors in selecting a nonvocal mode for language training with individual children are described, along with the results of this selection procedure as applied to five autistic children (5 to 8 yeas old). Attention is also directed to the rationale for, and the specific steps involved in, four techniques that parents can be trained to use to facilitate their child's learning and functional use of the nonspeech training mode. It is noted that in deciding which mode to use in training, the clinician should consider a child's functional communication channel and the characteristics of nonspeech modes. In addition to choosing a nonspeech mode, the use of the nonvocal mode with five autistic children required the training of language responses with one mode and assessing the children's learning of the mode, and training and assessing different language responses with the other mode. Training experiences with each of the four children are described, along with issues concerning long term training of the prescribed nonspeech mode. Illustrations are presented of how four techniques may be incidentally used to train speech (i.e., the Model Procedure, the Mand-Model Procedure, the Delay Procedure, and the Incidental Teaching Procedure). Appended materials include a list of 19 references, a description of experimental conditions, and a flowchart of a sample traning procedure. (SEW)